Public Preferences for Redistribution and Policy Outcomes: A Comparative Study. Chen Sharony, Shlomo Mizrahi, Miki Malul

Public Preferences for Redistribution and Policy Outcomes: A Comparative Study Chen Sharony, Shlomo Mizrahi, Miki Malul 1

Summary What are the factors affecting the gap between preference for income redistribution and policy? There is a mismatch between public preferences and policy in this field in certain countries. That is, in some countries the public displays a high demand for redistribution, but the government's social spending is low or vice versa. Ours is a comparative study which uses panel data from 24 OECD countries, from different years (1990-2012). Public preferences were measured by value surveys (ESS, WVS, ISSP) and policy is measured by social expenditure and the GINI index. The proposed sources for the gap between public preferences and policy are: social capital, ethnic heterogeneity, perceived government effectiveness and corruption. Results show that most countries have a small gap between public preferences and policy. A few countries studied (like Greece, Israel and Portugal) show a negative gap, where social spending is lower than the public preference. Other countries (like Sweden, Denmark and Luxemburg) show a positive gap, where public spending is higher than the public preference. Government effectiveness and corruption were found to be the main factors affecting the gap, but in some regressions social capital also has an effect on it. The influence of government effectiveness on the gap may mean that the public does not demand redistribution because faith in the government's ability to perform is lacking. Another interesting finding is that a "positive gap" was found in countries considered to have high government effectiveness and low corruption. This could mean that people believe there is too much redistribution in the country. 1. Introduction The study deals with public preferences concerning redistribution policy, and their effect on actual policy. Redistribution is defined in the literature as a reduction of inequalities in the distribution of wealth through government taxes and transfers (Durante & Putterman, 2009). There is a mismatch between public preferences and policy in this field in certain countries. That is, in some countries the public displays a high demand for redistribution policy, but the government's social spending is low or vice versa. This is a comparative study which examines the interaction between public preferences and policy in two dimensions: 1. Intentions-- Is the policy, as reflected in social spending measures, consonant with public preferences in the country? 2. 2

Outcomes Are policy outcomes, as reflected in income inequality measures, consonant with public preferences in the country? The second stage of the study attempts to assess the factors that shape the gap between these two dimensions. The proposed sources for the gap are: low social capital, high ethnic heterogeneity, a low level of perceived government effectiveness and a high level of perceived corruption. The study's main contribution moves toward a definition of the gap, using a measurable variable and clarifying the factors affecting it. The factors we propose were found in previous research to affect economic functioning or quality of government. Understanding the factors affecting the gap using international data for comparison, can lead to the construction of a policy which may reduce the gap and lead to civil awareness, political involvement, and to citizens voicing their demands for the policy they prefer. The research questions are: 1. Does actual policy reflect public preferences on the subject of the welfare state and redistribution policy? 2. What are the factors that explain the gap between public preferences for redistribution and policy outcomes? 2. Literature review 2.1 Preferences for Redistribution There are many studies concerning the factors affecting preferences for redistribution, such as: self-interest (Cusack, Iversen, & Rehm, 2008; Doherty, Gerber, & Green, 2006; Estevez-Abe, Iversen, & Soskice, 2001; Iversen & Soskice, 2001; Iversen & Soskice, 2006; Rehm, 2005; Rehm, Hacker, & Schlesinger, 2012); values and norms (Alesina & Glaeser, 2004; Alesina & Giuliano, 2009; Durante & Putterman, 2009); socialization (Svallfors, 1997), and culture (Alesina & Schuendeln, 2005; Andreß & Heien, 2001; Corneo & Grüner, 2002; Luttmer & Singhal, 2008; Sapienza, Zingales, & Guiso, 2006). Studies showed mixed findings about the relations between welfare state type and redistribution preferences (Arts & Gelissen, 2001; Gelissen, 2000; Jæger, 2006; Lipsmeyer & Nordstrom, 2003). Some studies found a connection 3

between welfare state type and redistribution preference, while others didn't find a connection or found a non-linear connection. 2.3 Public policy and Public opinion There is a body of literature addressing the topic of democratic deficit. A democratic deficit occurs when democratic institutions (particularly governments) fall short of fulfilling the principles of democracy in their practices. Theories of democratic deficit speak about discontent with the government as well as low voter turnout, low civic engagement and declining social capital, along with the claim that people hate politics. Russell Dalton claims that citizens in advanced industrial societies remain committed to democratic principles although they have gradually become more distrustful of politicians, detached from parties and doubtful about the public sector and institutions. The gap between aspiration and satisfaction is captured by the concept of democratic deficit (Dalton, 2004). Researchers claim there is a democratic deficit in the EU (Follesdal & Hix, 2006) and in the U.S. (Lax & Phillips, 2012). Other researchers maintain there is no democratic deficit: Norris shows that trust in politics and political institutions has not declined throughout the years and thus challenges the democratic deficit approach (Norris, 2011). Some researchers claim that the problem with the EU is not democratic deficit but a credibility crisis. If the EU could increase its credibility through more transparent decision-making and professionalism, then the public would accept its legitimacy and democratic deficit would disappear (Crombez, 2003; Majone, 1998). In sum, there is a claim in the literature about declining credibility and the performance of governments and democratic institutions which leads to decreased citizen involvement in politics. If the democratic deficit claim is true, this may explain the gap between public preferences and policy. On the other hand, many researchers maintain that there is no democratic deficit and the democratic institutions in Europe and the U.S. are functioning well. Another body of literature called "Agenda Setting" deals with the question: Does the government respond to public demands and does the public adjust its demands when there is a policy change? A normative model called the "Thermostat Model", which originates with Deutsch & Deutsch (1963) and Easton (1965), claims that the public 4

acts like a thermostat for politicians: 1. If there is a gap between public preferences and policy, the public will send a signal. 2. Politicians will respond to this signal with policy change. 3. As the policy comes close to the desirable level, the signal will weaken. Several studies found that the thermostat model works well. For instance, a study conducted in the U.S., Canada and the U.K. found high responsiveness of the public to policy change and high responsiveness of the government to public demands (Soroka & Wlezien, 2010). Wlezien (1995) found that the public responds to policy changes, and changes its preferences. However, politicians do not necessarily react to preferences changes by the public and therefore we see a certain circular trend in preferences over time. Baumgartner, et al. (2009) checked whether institutional differences between countries affect the level of politicians' responsiveness to public demands. They assumed that a gap between public preferences and outcomes will emerge and that this gap will be larger in the U.S. than in the parliamentary democracies (Belgium and Denmark). They found that despite institutional differences, the results were similar between the countries. Page & Shapiro (1983) found a match between public preference and policy in the years 1935-1979 in the U.S. They found higher matches on social issues (abortion, civil rights) than on welfare issues. They claimed that policy matches preferences, especially in subjects with high visibility. Soroka & Wlezien (2004) conducted research in the U.S., Canada and the U.K. They checked both the responsiveness of the public to policy changes and the responsiveness of the government to policy demand from the public. They found that the thermostat model works in each of the three countries. That is, in all three countries the public responds to a change in the level of spending and the level of spending is in line with the public demands. However, there are differences in the nature of the response that may be affected by the type of government structure. It was found that in the presidential system (the U.S.) responsiveness is higher than in the parliamentary system. In sum, most of the studies in the Agenda Setting field support the thermostat model that is, the government and the public respond to one another. However, it is important to mention that most of these studies were conducted in the U.S. and Canada. 5

2.2 Factors affecting the gap between public preferences and policy outcomes Our research assesses factors that may affect the gap between public preferences for redistribution and policy outcomes. The next section introduces the factors assumed to affect this gap and previous literature regarding them. 2.2.1 Social capital Social capital is defined as "the network of social connections that exists between people, and their shared values and norms of behavior, which enable and encourage mutually advantageous social cooperation." (Putnam, 1995). The term social capital was first introduced by Putnam, who investigated the functioning of local government in several Italian districts. He found that although they had the same budget, Northern districts had higher quality local government. He claimed that this was because Northern Italy had better civil engagement: higher involvement and sense of community, higher egalitarian and democratic ethos. That is, social capital is a measure of the level of civil involvement in country or community (Putnam, Leonardy, & Nannety, 1993). Much research deals with the relation between social capital and democracy and societal functioning (Brass, Butterfield, & Skaggs, 1998; Fukuyama, 1995; Newton, 1997; Porta, Lopez-De-Silane, Shleifer, & Vishny, 1996; Putnam et al., 1993; Putnam, 1995; Putnam, 2002; Putnam, 2000). Fukuyama (1995) claimed that a high level of trust and social capital enables better functioning of a country's institutions. Porta et al. (1996) found that high level of trust brings to better economic performance of the country. Countries with a high level of social capital will show civic participation, which leads to extensive government oversight. This could be manifested through non-profit organizations in the relevant field, providing information and supervising decisionmaking and policy. So we shall also expect a link between public opinion and policy outcomes. Therefore, the first hypothesis is: H1: Countries with lower social capital will show a bigger gap between public preference and actual policy than country with higher social capital. 6

Social capital was not examined in previous literature as a factor affecting the gap between public preferences and policy. Although there is much research in this field, it mainly concerns the effect on political participation, quality of government and health. If we find such an effect, it will be another application of the positive effect of social capital on societies. 2.2.2 Ethnic heterogeneity Much as in the case of social capital, where low levels of trust between citizens sometime obtains, we would expect ethnically diverse countries to show more suspicion between the different groups and therefore less support for universal policy. In this situation we would also expect to have a gap between public preference and policy. This can manifest itself, for example, by a low income population characterizing a certain minority (as in Israel where most of the low income populations are Arabs and Ultra-Orthodox Jews). This situation may lead to diminished support for redistribution by the majority. Another factor operating in ethnically diverse countries is political division, which leads to more interest groups and higher costs of policy. Studies found that ethnically diverse societies tend to show lower public spending (Alesina, Baqir, & Easterly, 1999; Kuijs, 2000; Luttmer, 2001) and provide fewer public services (Easterly & Levine, 1997). La Porta, Lopez-de-Silanes, Shleifer, & Vishny (1999) found that countries with a high level of ethnic heterogeneity had lower government quality. Alesina et al. (2001) argued that ethnic heterogeneity is the main reason why the U.S. does not have a European style welfare state. Kuijs (2000) also found that in ethnically diverse societies the quality of public spending was lower. This means that the problem is not only one of public choice but also of technical efficiency. On the other hand, Annett (2001) found that ethnically diverse countries showed higher public spending (in total). He claimed that these countries have lower political stability, which leads to higher government consumption, causing coalition governments to make efforts to fend off opposition demands. Easterly & Levine (1997) claimed that ethnically diverse countries have more interest groups, which leads to less than optimal public decision-making. Another study found that in Latin 7

America an increase in ethnic diversity leads to increased political division, manifested by a large number of political parties (Birnir & Van Cott, 2007). Gilens (2009) claimed that in reality the American public supports many components of the welfare state. That is, they believe that government should spend more on education, health etc. The source of the resistance to a more generous welfare policy is the perception that all the money goes to African-Americans and the belief that they are lazy. He points to media coverage of the welfare subject as the main source for that perception. There is probably a relation between ethnic heterogeneity and social capital. Research has found that ethnically homogeneous countries have high social capital, meaning ethnic heterogeneity may have a negative effect on social capital (Alesina, Gleaser & Sacredote, 2001; Alesina & La Ferrara, 2002; Coffé & Geys, 2006; Letki, 2008; Putnam, 2007). In sum, ethnic heterogeneity was found to affect the quality of government and economic growth in previous research via the mechanism of mistrust between the different groups and the resultant political division. In divided societies, where there is a feeling that not all groups are contributing equally or that there is a group which is the main beneficiary of welfare policy, there may be a gap between preferences and actual support for the policy. Therefore, our second hypothesis is: H2: Countries with high ethnic heterogeneity will show a bigger gap between public preferences and policy outcomes than countries with low ethnic heterogeneity. 2.2.3. Perceived government effectiveness and corruption The following hypothesis deals with perceived government effectiveness and corruption. These two factors concern the public's trust in the government. Scholz & Lubell (1998) found that a high level of trust in government leads to a higher rate of tax payments. Oh & Hong (2012) claimed that a low level of trust leads to low willingness to pay taxes, which in turn leads to low support for redistribution. Algan, Cahuc, & Sangnier (2011) found that a medium level of trust in government leads to reduced welfare policy, as in the Anglo-Saxon countries and Japan. 8

Yamamura (2012) examined the relations between trust in government and support for redistribution in Japan. He found that: 1.In regions where trust in local government is high people tend to show more support for redistribution. 2. In regions where trust in government is high, the tax burden is perceived as lower. Roosma, Gelissen, & van Oorschot (2012) found that in European countries there is a positive perception about the goals of the welfare state and yet, at the same time, there is criticism of the efficiency, effectiveness, and results of the welfare state policies. This difference is especially true for Eastern and Southern European countries, whereas in Northern and Western European countries there is less criticism of efficiency and the effectiveness of policy results. This finding indicates that there can be support in principle for redistribution, but in practice such a policy encounters resistance due to criticism of the government s effectiveness. On the other hand, research in Sweden, Norway and the U.S. found no relation between trust in government/perceived government effectiveness and preferences for welfare policy (Edlund, 1999). Svallfors (1999) found that institutional quality (not that of politicians or political parties) and trust in institutions affects the level of support for welfare policy. In countries with a high level of trust and a low level of corruption, people who support welfare policy in principal will tend to support welfare policy in practice (higher taxes, increased benefits). In addition, corrupt governments may exhibit ineffective policy, and therefore policy outcomes may not reflect public preferences. Brewer, Choi, & Walker (2007) found that better control of corruption leads to better governance. That is, when corruption is low, budgets are being used more efficiently. Olken (2006) examined the effect of corruption on wealth in an anti-poverty program in Indonesia, which distributed subsidized rice to poor households. He estimated that the welfare loss from the corruption was large enough to offset the potential welfare gains from the plan. That is, corruption has an effect on welfare plans' effectiveness. Kuijs (2000) found that corruption affects the level of public social spending on health and education that is, less corrupt governments spend more but it doesn't have an effect on policy outcomes, e.g. on technical efficiency. 9

There may be a high correlation between ethnic heterogeneity and corruption. Studies found that corrupt countries are more ethnically diverse (La Porta et al., 1999; Glaeser & Saks, 2006). In sum, previous research dealing with trust in government found it to affect willingness to pay taxes and support for redistribution. We believe, with Roosma, et al. (2012), that when there is low trust in government or low perceived government effectiveness, people who support redistribution in principle may not support it in practice, because they don t believe in government's ability to provide this policy effectively. In addition, corruption can affect welfare plans' effectiveness as well as government's effectiveness in general. If the public feels that money is being wasted, this could create a gap between public preference and what it demands from the government. Therefore, our third hypothesis is: H3: Countries with low level of perceived government effectiveness and high level of perceived corruption will show a bigger gap between public preferences and policy outcomes than countries with high perceived government effectiveness and low level of perceived corruption. 2.3 Prior research A few researchers have investigated the relations between preferences and actual policy with regards to redistribution: Kuhn (2012) found that in OECD countries the policy indices: change in GINI before and after taxes, social spending index and GINI post transfers and taxes, showed a strong and significant relation to preference for redistribution. That is, in countries with high support for redistribution, there is also a high level of redistribution, coupled with a high level of social spending and low income inequality. In addition, Kuhn found a relation between individual preference for redistribution and wider political perception. He also found that the relation between individual preference and wider political perception becomes non-significant when countries' national characteristics, such as political institutions or ethnic diversity are added to the equation. His conclusion is that specific characteristics of a nation, like its political institutions and ethnic diversity, affect both individual citizens' preferences for redistribution and policy outcomes. 10

Jæger (2006) investigated the relations between welfare state type and support for redistribution in Western European countries. He found a correlation between the level of public spending and support for redistribution. In addition, he found a nonlinear relation between welfare state type and preference for redistribution. This could mean that people from social-democratic countries, for example, may feel that there is enough redistribution in the country, and therefore oppose more of it. It can be noted that most previous research which looked at this topic found a match between preference and policy in most of the European and OECD countries (including the "Agenda-Setting" literature). However, our suggested method is somewhat different, as is the set of countries we examine. The present study considers the relation as not necessarily linear, as claimed by Jæger (2006). For instance, it is possible that in the Scandinavian countries there is a feeling that too much redistribution has already occurred and therefore there is an objection to more redistribution. Illustration No. 1: The research model Policy outcome s Policy intention s Social capital Gap between public preferences and policy Ethnic heterogeneity Public preference Perceived government effectiveness and corruption 3. Methodology The study has 2 phases: 1. Measuring the gap between public preference and actual policy. 2. Panel regressions, assessing the factors affecting this gap. 11

We calculated 2 kinds of gaps: 1. between preference and policy intentions (measured by social spending). 2. between preference and policy outcomes (measured by the level of income inequality). 3.1 Data The study uses data from 24 OECD countries from different years (between 1990 and 2012).These 24 countries were chosen because they have all the data needed for the regressions. Public preference was measured by 3 surveys: o European Social Survey (ESS) o World Value Survey (WVS) o International Social Survey Program (ISSP). Support for redistribution was measured by the question: "Should governments reduce income differences/should incomes be made more equal?" (Depending on the survey). Policy outcomes were measured by income inequality measures: - GINI index after taxes and transfers (OECD data). - MGINI - a modified GINI index that takes into account the moderating effect of in-kind government benefits. Thus there may be a situation in which a country has a high level of income inequality but a generous welfare system. This country will show lower MGINI, and will be considered more egalitarian (Malul, Shapira, & Shoham, 2013). - GINI difference- the difference between GINI index before and after taxes and transfers. 1-(GINI_post_transfers/GINI_before_transfers). Policy intentions: - OECD social spending indexes: o Government Social Spending, % of GDP o Government Social Spending per head 12

3.2 Statistical methods First step: Measuring the gap: We measured the gap using 2 different techniques: 1. Difference index: 1. We created a 5 years average for each parameter (with a separate index for preference and policy). The 15 year database was divided into three groups: 1990-1998, 1999-2005, 2006-2012. 2. We calculated the index between all countries for the same period {valuemin)/(max-min)}. 3. We subtracted preference index from policy index. When the difference index is close to 0 there is a match between public preference and policy measure. When the difference index is high there is a positive gap (the government spends more than the public demands). When the difference index is low there is a negative gap (the government spends less than the public demands). 2. Ratio: Dividing policy measure (value) by preference measure (value). Second step: Measuring the factors affecting the gap The relations between the factors were examined using panel linear regression, which takes into account the variance between countries and between the years within each country. The dependent variable was the gap, defined in the first stage. The independent variables were: -Social capital: measured by a survey question: "Most people can be trusted or you can't be too careful" taken from value surveys. The question was chosen because it represents the level of trust in one's environment, which represents the level of social capital. Previous research also used this question to measure social capital. - Ethnic heterogeneity: measured by the ethnic fractionalization index reflects the probability that two randomly selected people from a given country will not share a certain characteristic (ethnicity). The higher the number the less probability of those two people sharing that characteristic ( Alesina, Devleeschauwer, Easterly, Kurlat, & Wacziarg, 2003). - Government effectiveness: 2 different measures are used: 13

1. ICRG International country risk guide - ICRG collects political information and financial and economic data, converting these into risk points. Higher values indicate higher quality of government. The measure for government effectiveness is comprised of the variables: "Corruption", "Law and order" and "Bureaucratic quality". 2. GEE (World bank) - These indicators are based on several hundred individual variables measuring perceptions of governance, drawn from 31 separate data sources constructed by 25 different organizations. Government effectiveness combines into single grouping responses on the quality of civil servants, the independence of the civil service from political pressures, and the credibility of the government's commitment to policies. The main focus of this index is on "inputs" required for the government to be able to produce and implement good policies and deliver public goods. -Corruption CPI (corruption perception index)- is derived by "Transparency International". CPI focuses on corruption in the public sector and defines corruption as the abuse of public office for private gain. The CPI score relates to perceptions of the degree of corruption as seen by business people, risk analysts and the general public and ranges between 10 (highly clean) and 0 (highly corrupt). Since there was high correlation between some of the indices 1 and since the ICRG index includes both government effectiveness and corruption, we ran a basic regression (model 1) with three independent variables: ethnic heterogeneity, social capital and government effectiveness measured by ICRG. In addition, we ran regressions where government effectiveness and corruption were entered separately. Model 2 included ethnic heterogeneity, social capital and government effectiveness (measured by GEE). Model 3 included ethnic heterogeneity, social capital and corruption (measured by CPI). We conducted a Hausman test for each of the regressions and found that in all the models based on difference index, there is no endogeneity and used "random effects". In addition, for most of the ratio index regressions there is endogeneity (besides MGINI), and therefore used "fixed effects". 1 The correlation between ICRG (government effectiveness and corruption) and CPI (corruption) was 0.848. The correlation between GEE (government effectiveness) and CPI (corruption) was 0.858. These variables were entered in separate regressions. The correlation of social capital with effectiveness and corruption was also high, but lower: 0.809 with ICRG, 0.758 with GEE and 0.749 with CPI. These variables were entered into the same regression. 14

We believe there is no endogeneity in the model. The existence of the gap is not supposed to affect social capital: social capital is probably affected by cultural characteristics of the country, so it shouldn t change much throughout the years. Even if there are changes in social capital, they would probably be affected by major economic and social transitions and not by preferences for redistribution or by the gap between preference and policy. The gap may affect the level of corruption and government effectiveness, but not immediately. This effect can happen after a few years. For example, if the level of corruption is high and leads to a gap, the public may demand to narrow this gap, but this doesn't happen immediately. Finally, ethnic heterogeneity is an exogenous variable that cannot be affected by the gap. Ordinal and binary regressions: An issue that came up during the study is the difference between a positive gap (when the country spends more than the public demands) and a negative gap (when the country spends less than the public demands). In the linear regression there is a linearity assumption. That is, that positive gap is the optimal situation, no gap lies in the middle and a negative gap is the least preferred situation; while, at least theoretically, our assumption is that the best situation is no gap. To address this issue, we conducted two regression types: 1. Ordinal logit- which differentiates between the three groups (positive gap, small or no gap, negative gap). The cut-off for the groups was 0.1 and -0.1 in the difference index. This created 3 groups for each variable, generally at the same size (for "spending per head" the cut-off was changed to 0.05 and -0.2 for the groups to be at the same size). 2. Binary logit- based on the ordinal logit but separating to 1- gap (positive and negative) and 0 no gap. 4. Results 4.1 Measuring the gap The graphs shown here are all based on the difference index. The ratio index shows similar results. The difference index was calculated for three different periods: 1990-1998, 1999-2005, 2006-2012. Illustrations nos. 2-4 shows the different indices between preference and social spending as a percent of GDP for all periods. 15

Denmark Sweden Germany Belgium Italy France Luxemburg Austria Netherlands Finland Japan Norway New-Zealand United Kingdom United States Canada Australia Spain Switzerland Iceland Ireland Portugal Greece Israel Denmark Sweden Belgium Netherlands Finland Canada France Germany New-Zealand Austria Norway United States Australia United Kingdom Iceland Spain Italy Japan Switzerland Ireland Israel Portugal Illustration no. 2: Difference index of the gap between public preference and social spending as % of GDP, averaged for the years 1990-1998 1 0.8 0.6 0.4 0.2 0 Illustration no. 3: Difference index of the gap between public preference and social spending as % of GDP, averaged for the years 1999-2005 -0.2-0.4-0.6-0.8-1 1 0.8 0.6 0.4 0.2 0-0.2-0.4-0.6-0.8 16

Denmark Italy Luxemburg New-Zealand Sweden Belgium France Germany Netherlands Canada Finland United Kingdom Japan Austria Norway United States Switzerland Australia Spain Portugal Ireland Greece Iceland Israel Illustration no. 4: Difference index of the gap between public preference and social spending as % of GDP, averaged for the years 2006-2012 1 0.8 0.6 0.4 0.2 0-0.2-0.4-0.6-0.8-1 -1.2 Most of the countries were close to 0, which means no gap or a small gap. Positive gap (high social spending with low support for redistribution) was found mainly in Denmark and Sweden. Negative gap (low social spending with high support for redistribution) was found mainly in Israel, Greece and Portugal. We can also see that in Israel the gap has been widening throughout the years. For inequality measures, the direction of the different measures is the other way around. That is, a high score means a negative gap and a low score means a positive gap. Illustrations no. 5-7 shows the different indices between preferences and income inequality measures for all periods. 17

Portugal Israel Spain Greece United Kingdom France Ireland United States New-Zealand Italy Austria Finland Belgium Japan Canada Australia Norway Iceland Germany Netherlands Luxemburg Sweden Denmark Illustration no. 5: Difference index of the gap between public preference and GINI index post transfers and taxes, averaged for the years 1990-1998 1 0.8 0.6 0.4 0.2 0-0.2-0.4-0.6-0.8-1 -1.2 Illustration no. 6: Difference index of the gap between public preference and GINI index post transfers and taxes, averaged for the years 1999-2005 1.5 1 0.5 0-0.5-1 -1.5 18

Israel Portugal Greece Spain United States France Ireland United Kingdom Austria Japan Germany Australia Iceland Finland Belgium Switzerland Sweden Netherlands Italy Canada New-Zealand Norway Luxemburg Denmark Illustration no. 7: Difference index of the gap between public preference and GINI index post transfers and taxes, averaged for the years 2006-2012 1 0.8 0.6 0.4 0.2 0-0.2-0.4-0.6-0.8-1 Here again, most of the countries had an index close to 0, which means no gap or a small gap. Positive gaps (low inequality with low support for redistribution) were found, once again, mainly in Denmark, Sweden and Luxemburg. Negative gaps were found, again, mainly in Israel, Portugal, Greece and Spain. 4.2 Measuring the factors affecting the gap All results shown from this stage are based on the ratio index. Results from the difference index were similar. Panel regression results for social expenditure measures are shown in table 1. 19

Theoretical parameter Table 1: Panel linear regression results factors affecting the gap between preference and level of social spending Independent Variable Dependent variable: ratio between public preference and social spending per head Model 1 Model 2 Model 3 486.69 618.93 492.93 Dependent variable: ratio between public preference and social spending as % of GDP Model 1 Model 2 Model 3 Ethnic Ethnic -0.89-0.84-0.90 Heterogeneity Fractionalization (605.97) (672.18) (637.93) (1.39) (1.36) (1.38) Social Capital Most people can be trusted 9.44 (12.75) 34.02*** (11.86) 20.19 (13.09) -0.0018 (0.02) Government 2316.37** 2.48* Effectiveness ICRG and Corruption (938.53) (1.5) 0.01 (0.02) Government -327.6* 0.33 GEE Effectiveness (167.78) (0.24) Corruption CPI Constant -1083.09 (719.49) 91.4 (645.10) 7.33 (7.33) -264.90 (651.06) 2.5* (1.46) 4.32 (1.14) 0.008 (0.02) 0.002 (0.01) 3.88*** (1.22) Adjusted R 2 0.384 0.138 0.328 0.187 0.009 0.144 N 67 67 67 91 91 91 Wald chi2 13.26 9.61 standard deviations are reported in parentheses *statistically significant at the 10% level. **statistically significant at the 5% level. ***statistically significant at the 1% level. 7.47 3.44 2.60 0.82 The main factor affecting the gap, as can be seen in Table 1, is ICRG, which represents government effectiveness and corruption. It has a positive effect on both measures of social spending, which means countries with more effective and less corrupt government shows smaller gap or positive gap. In model 2, which uses the GEE measure for government effectiveness, government effectiveness had a negative and significant effect, which contradicts our hypotheses. In addition, in model 2 social capital becomes positive and significant, meaning that higher social capital leads to a positive gap. For the gap between preference and social spending as % of GDP, none of the factors was significant in models 2 and 3. Panel regression results for inequality measures are shown in table 2. 20

Theoretical parameter Table 2: Panel linear regression results factors affecting the gap between preference and level of income inequality Independent Variable Dependent variable: ratio between public preference and GINI index Dependent variable: ratio between public preference and GINI difference Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 Ethnic Ethnic -7.48-8.19-12.06 0.005 0.002 0.005 Heterogeneity Fractionalization (16.44) (17.07) (16.06) (0.25) (0.02) (0.02) Social Capital Most people can -0.47-0.68** -0.57* -0.00001 0.0004 be trusted (0.35) (0.34) (0.34) (0.0004) (0.0004) Government -80.04*** 0.097*** Effectiveness ICRG and Corruption (27.02) (0.03) Government -13.62** 0.005 GEE Effectiveness (6.18) (0.007) Corruption CPI Constant 163.26*** (19.63) 128.16*** (16.3) -0.54*** (0.20) 142.19*** (15.98) -0.01 (0.03) 0.04 (0.02) 0.0003 (0.0004) 0.0003 (0.0002) 0.031 (0.02) Adjusted R 2 0.573 0.562 0.593 0.435 0.409 0.413 N 59 59 59 54 54 54 Wald chi2 25.86 20.63 standard deviations are reported in parentheses *statistically significant at the 10% level. **statistically significant at the 5% level. ***statistically significant at the 1% level. 26.09 10.87 2.60 3.21 The main factor affecting the gap is, again, ICRG (the only significant factor in model 1). It affects negatively the gap between preference and GINI and positively the gap between preference and GINI difference, consonant with our hypotheses. This means that countries with high government effectiveness and low corruption tend to show positive gap. If we look at models 2 and 3 we can see that government effectiveness, corruption and social capital all have positive and significant effect on the gap. This means that, as we hypothesized, countries with higher government effectiveness, lower corruption and higher social capital show small gap or positive gap. For the GINI difference measure models 2 and 3 are not significant. Linear regressions showed that the main factor affecting the gap is government effectiveness and corruption. When separating government effectiveness and corruption to different models (models 2 and 3), social capital was also found to affect the gap. 21

Ordinal regressions In ordinal regressions there is a gross distinction between the three groups: countries with a positive gap received the value 1. Countries with no gap or a small gap received the value 0 and countries with a negative gap received the value -1. The results for the gap between preference and social spending were very similar to the linear regression results. That is, government effectiveness and corruption showed a positive and significant effect. For the gap between preference and inequality measures none of the factors were significant in the first model, but in model 2 and 3 both government effectiveness and corruption were negative and significant, consonant with the hypothesis. This means that countries with high government effectiveness and low corruption show small gaps between public preferences and policy (the full tables are reported in Appendix 1). In conclusion, Ordinal regressions results show that by looking at gross distinction between the groups we achieve similar results, which strengthens our findings about the effect of government effectiveness and corruption on the gap. Binary regressions In this regression we created binary factor where 1=gap and 0=no gap. The regression results are presented in Tables 3 and 4. 22

Table 3: Logit regression results factors affecting the gap between preference and level of social spending Theoretical parameter Independent Variable Dependent variable: ratio between public preference and social spending per head Model 1 Model 2 Model 3 1.20 4.81 2.45 Dependent variable: ratio between public preference and social spending as % of GDP Model 1 Model 2 Model 3 0.17 1.54 0.84 Ethnic Ethnic Heterogeneity Fractionalization (3.88) (5.03) (4.10) (1.79) (1.90) (1.90) Social Capital Most people can be 0.07 0.09 0.06 0.11* 0.05 trusted (0.11) (0.13) (0.12) (0.06) (0.05) Government -19.27* -18.90*** Effectiveness ICRG and Corruption (1.77) (6.61) Government -6.39* -2.58** GEE Effectiveness (3.38) (1.30) Corruption CPI Constant N Wald chi2 14.89* (8.51) 67 3.65 8.84 (5.81) 67 3.99-0.13 (0.09) 7.73 (6.65) 67 11.89*** (4.14 91 2.77 (2.16) 91 0.07 (0.06) -0.10** standard deviations are reported in parentheses *statistically significant at the 10% level. **statistically significant at the 5% level. ***statistically significant at the 1% level. In Table 3, quite similar to previous regressions, government effectiveness and corruption measures are negative and significant in most of the models, which is consonant with the hypothesis. The other factors are not significant. This means that the higher the government effectiveness and the lower the corruption, the lower the probability for a gap. We can also see that social capital had a positive effect on the 3.16 gap, contrary to our expectation, but this result is not significant. 8.67 4.77 (0.05) 5.40** (2.73) 91 4.71 23

Table 4: Logit regression results factors affecting the gap between preference and level of income inequality Theoretical parameter Independent Variable Dependent variable: ratio between public preference and GINI index Model 1 Model 2 Model 3 Ethnic Ethnic -0.05 0.23-1.47 Heterogeneity Fractionalization (3.37) (3.41) (2.38) Social Capital Most people can 0.06 0.08 be trusted (0.09) (0.09) Government -13.26 Effectiveness ICRG and Corruption (8.69) Government -4.33* Gee Effectiveness (2.29) Corruption CPI Constant N Wald chi2 standard deviations are reported in parentheses 10.19* (5.47) 59 2.79 4.60 (3.48) 52 3.98 *statistically significant at the 10% level. **statistically significant at the 5% level. ***statistically significant at the 1% level. -0.01 (0.05) -0.16* (0.08) 9.33* (4.89) In the regressions for the gap between public preferences and income inequality measures (Table 4), it can be seen that in the basic model none of the factors had significant effect. In models 2 and 3 both government effectiveness and corruption had significant effect, consistent with our hypothesis: Meaning that countries with high government effectiveness and low corruption have a lower probability of showing a gap between public preference and policy outcomes. It is important to note that logit regressions, measuring the effect without distinction between types of gap, found that government effectiveness and corruption were the main factors affecting the gap. This means that when we distinguish between countries with a gap and countries with no gap, we find an effect of the same factors as found in the linear and ordinal regressions (government effectiveness and corruption). The lower the government effectiveness and the higher the corruption, the higher the probability of having a gap. This result could mean that in countries with a negative gap these factors are especially strong (government effectiveness is especially low and corruption is especially high) so they are "dragging down" the countries with positive gap. 52 3.75 24

5. Discussion This study examined the relation between public preferences for redistribution and policy intentions (measured by social spending) and outcomes (measured by GINI index) in OECD countries for different years. The first stage of the study defined the gap. The results from this stage show that countries like Israel, Portugal, Greece and Spain show mostly a negative gap (spending less than the public demands or are less egalitarian than the public wishes). Countries like Denmark, Sweden and Luxemburg show mostly positive gaps (spending more than the public demands or are more egalitarian than the public wishes). Countries like Austria, Belgium, U.S.A. and Canada mostly show a match between public preferences and policy or a small gap. The second stage of the study looked at the factors affecting this gap. The results show that among the factors suggested in the study, government effectiveness and corruption were the most significant elements affecting the gap between public preference and policy. This finding supports hypothesis 3. Another factor that was found to be significant in some regressions was social capital- which partly supports hypothesis 2. Ethnic heterogeneity wasn't found to affect the gap, contrary to the first hypothesis. The issue of consistency between public preference and policy is one of the aspects of democratic functioning. This is an even more significant test of democracy's dealing with the issue of redistribution, since this issue normally gets little coverage in the media. Page & Shapiro (1983) claimed that when visibility of certain subjects is low, the probability of government responding to public demand is similarly lower. The findings for U.S.A., U.K., Canada and other countries support the Agenda Setting theory (Baumgartner et al., 2009; Page & Shapiro, 1983; Soroka & Wlezien, 2004; Soroka & Wlezien, 2010; Wlezien, 1995). In these countries, there is a match between public preferences and actual policy, much like the predictions of the Agenda Setting theory and the Thermostat model. The match found in the U.S. can be explained by the type of governance, as Soroka & Wlezien (2004) claimed. This means that in a presidential system (the U.S.), government responsiveness is greater than in the parliamentarian system. This can also be explained by the level of centralization. Under centralized governance (where the central government is strong) responsiveness to public demands will be higher than under federal governance (where local states has more power) (Soroka & Wlezien, 2004). 25

In contrast to previous Agenda Setting research, we also found countries with no match between public preferences and policy. First, we will discuss countries with a negative gap (lower level of spending than the public demands and higher level of inequality than the public expects): Israel, Greece, Portugal and sometimes Spain. One quality these countries (except Israel) have in common is that they have been linked as belonging to the so-called PIGS countries (Portugal, Italy, Greece and Spain): These are EU member countries that were unable to refinance their government debt or to bail out over-indebted banks on their own during the debt crisis, beginning in 2008. It is possible that the gap or the low government effectiveness and high corruption, were the factors that lead to economic dysfunction or the other way around. This can be a topic for future research. Another interesting finding is that some countries show a positive gap (higher level of public spending than the public demands and lower level of inequality than the public expects). These are social-democratic countries, considered to have a good quality of government, like Denmark, Sweden and Luxemburg. This raises the question: Is this a situation where the government is not in touch with public demands? Understood in thermostat model terms, the public demands lowering the level of spending and the government has not responded. Another possibility is that the public feels there is too much redistribution in these countries, meaning there is a non-linear relation between welfare state type and preference for redistribution (Jæger, 2006). Yet another option is that in these countries the government is confident enough to lead a policy even when there is no demand from the public, assuming that the public will adopt governmental preferences. The main finding from the second stage of the study is that government effectiveness and corruption are the main factors affecting the gap. This finding is consistent in both linear regression (when a higher score equates to a positive gap and a lower score to a negative gap), Ordinal regression (-1 for negative gap, 0 for small gap, 1 for positive gap) and logit regression (0 for small gap, 1 for negative or positive gap). This could mean that these factors (government effectiveness and corruption) are especially strong in countries with a negative gap. The meaning of this finding may be that the public doesn't demand the policy it believes in because of lack of trust in the government's ability to deliver the policy effectively, or because it believes the government is corrupt. 26

Another factor found to affect the gap in some regressions was social capital- the level of trust in other people. This could mean that countries with low social capital show a low level of civic action, and the citizens don t act to demand the policy they want or to supervise the government. It is also important to mention that ethnic heterogeneity was not found significant in any of the regressions. That is, contrary to former research that found an effect of ethnic heterogeneity on quality of government and support for redistribution (Alesina et al., 1999; Alesina et al., 2001; Easterly & Levine, 1997; Kuijs, 2000; Luttmer, 2001), in this study it had no effect on the gap between preference and policy. Another interesting point is that government effectiveness and corruption had a significant effect on both policy intentions (public spending) and policy outcomes (income inequality). That is, these factors bring about low trust in government, which may in turn lead to a failure to demand the policy citizens want. They can also lead to ineffective performance, even if the intention is to deliver the desired policy. It seems that improving government effectiveness and fighting corruption may contribute to reducing the gap between public preference for redistribution and policy, and can lead to better functioning democracy which acts according to the citizens' preferences. 27

References Alesina, A. Devleeschauwer, A., Easterly, W., Kurlat, S., & Wacziarg, R. (2003). Fractionalization. Journal of Economic Growth, 8(2), 155-194. Alesina, A., & Glaeser, E. L. (2004). Fighting poverty in the US and Europe: A world of difference. Oxford : Oxford University Press. Alesina, A., & La Ferrara, E. (2002). Who trusts others?. Journal of Public Economics, 85(2), 207-234. Alesina, A., & Schuendeln, N. F. (2005). Good bye Lenin (or not?): The effect of communism on people's preferences. National Bureau of Economic Research Working Paper Series, No.w11700. Retrieved from: http://www.nber.org/papers/w11700 Alesina, A. F., & Giuliano, P. (2009). Preferences for redistribution. National Bureau of Economic Research Working Paper Series, No. 14825. Retrieved from: http://www.nber.org/papers/w14825; http://www.nber.org/papers/w14825.pdf Alesina, A., Baqir, R., & Easterly, W. (1999). Public goods and ethnic divisions. The Quarterly Journal of Economics, 114(4), 1243-1284. doi:10.1162/003355399556269 Alesina, A., Glaeser, E., & Sacerdote, B. (2001). Why doesn't the US have a European-style welfare system?. National Bureau of Economic Research Working Paper Series, No. 8524. Retrieved from http://www.nber.org/papers/w8524; http://www.nber.org/papers/w8524.pdf Algan, Y., Cahuc, P., & Sangnier, M.,. (2011). Efficient and inefficient welfare states. IZA Discussion No. Paper No. 5445. Andreß, H., & Heien, T. (2001). Four worlds of welfare state attitudes? A comparison of Germany, Norway, and the United States. European Sociological Review, 17(4), 337-356. Annett, A. (2001). Social fractionalization, political instability, and the size of government. IMF Staff Papers 561-592. Arts, W., & Gelissen, J. (2001). Welfare states, solidarity and justice principles: Does the type really matter? Acta Sociologica, 44(4), 283-299. Baumgartner, F. R., Breunig, C., Green Pedersen, C., Jones, B. D., Mortensen, P. B., Nuytemans, M., et al. (2009). Punctuated equilibrium in comparative perspective. American Journal of Political Science, 53(3), 603-620. Birnir, J. K., & Van Cott, D. L. (2007). Disunity in diversity: Party system fragmentation and the dynamic effect of ethnic heterogeneity on Latin American legislatures. Latin American Research Review, 42(1), 99-125. Brass, D. J., Butterfield, K. D., & Skaggs, B. C. (1998). Relationships and unethical behavior : A social network perspective. Academy of Management Review, 23(1), 14-31. doi:10.5465/amr.1998.192955 Brewer, G. A., Choi, Y., & Walker, R. M. (2007). Accountability, corruption and government effectiveness in Asia: An exploration of World Bank governance indicators. International Public Management Review, 8(2), 200-217. 28